The analysis of this case has progressed and expanded. A general overview of the major issues surrounding this case can be found on the LandisCase Wiki, where the Internet analysis of the Floyd Landis case are being summarized.

[I've found a key study that in some ways supports the ideas in this article and in other ways doesn't, but raises more disturbing issues fundamental to the test's reliability. You can read all about it here), although you probably still want to read this article for some of the background.]

Floyd's Exogenous Testosterone

Let me get straight to the point: it's impossible to tell for sure that anyone has taken synthetic testosterone.

Unfortunately, the way Floyd Landis' exogenous testosterone test has been portrayed in the media is as if it were a perfectly definitive test. Like pink for pregnant and white for not (not really a good example, since that isn't so accurate). Such tests do exist: tests with a binary outcome, yes or no, and an extremely low false positive or false negative rate. This is simply not one of them.

There is no difference between synthetic testosterone and naturally produced testosterone - they're one and the same chemical. Same atoms, in the same configuration, forming the exact same molecule, with identical chemcial properties. At least at the atomic level. Once you mix natural and synthetic testosterone, you can't separate them again, any more than you could separate Evian from Poland Springs bottled water after they'd been mixed. Actually that's a bad example. It would be more akin to separating two kinds of distilled water from each other. Even that would be easier than testosterone, since one would presume that distilled water sources don't change rapidly.

At any rate, natural and synthetic testosterone are usually different at the subatomic level. All the carbon in the world has six protons, and almost all the carbon in the world has six neutrons (called carbon-12). Some small portion of the carbon though, has seven neutrons (carbon-13), and an even smaller portion has eight (carbon-14).

Carbon-13 reacts chemically just like carbon-12. Except that some chemical processes prefer, just slightly, to use carbon-12 atoms instead of carbon-13 atoms. This means that the resulting molecules will have less carbon-13 than the average amount we find on earth. This difference is sometimes called d13C, d for delta (difference) 13C for carbon-13. This difference is generally expressed in "per mil" which is like percent only thousandths instead of hundredths (the symbol is "⁰/₀₀" instead of "⁰/₀"). The value of d13C is the difference in parts per thousand from the average earth-wide carbon-13 concentration. A d13C of -25 means that a sample has 25 less parts of carbon-13 per thousand than the earth-wide average. [1] Clarification - this is parts per thousand out of the expected carbon-13, not out of the total carbon. If you expect an average sample of some chemical to have 1000 carbon-13 atoms, then a sample with a d13C of -25 would only have 975 carbon-13 atoms.

A typical value for a person's testosterone appears to be anywhere from -21 to -28 per mil (studies I've found don't agree on this) [2, 3]. Synthetic testosterone's d13C is -26 to -30 per mil. [3] It's manufactured from soy, which apparently has a similarly low carbon-13 ratio.

When an athlete takes synthetic testosterone, it mixes with their natural testosterone in their body, and the resulting d13C is a weighted average of the two. In one study five subjects who had a range of testosterone d13C values between -26 and -28 were given 250mg of testosterone each; afterwards their testosterone d13C values were found to be in the range of -29 to -30. [2] Note that there seems to be some discrepancy in these study values; either 250 mg is an overwhelming amount of testosterone compared to what's in the body, or the testosterone used in this study has an atypical d13C value.

You ought to be able to see now how it's impossible to really identify exogenous testosterone. It's possible that a person with a low normal d13C value could dope, and still have a lower d13C value than another person with a high normal value, who does not dope. To avoid these problems, the test used on the athletes compares the d13C value of their testosterone with another natural steroidal compound in the body. The values should be similar -- although they aren't expected to be completely identical [2]; I'm not sure why this is, but it is one of serveral questions that's of fairly critical importance.

Testosterone's natural d13C value is within a broad range, because it depends on what we eat. [6, 7] The foods we eat typically have d13C values anywhere from -10 to -30 (always negative, because photosynthesis tends to prefer carbon-12 over carbon-13). This is actually two distinct ranges, with most plants in the range of -22 to -30, and a small number of plants (including corn and sugar cane) coming in at -10 to -14, because their photosynthesis process is slightly different. [1]

The most important question I have in all of this is: how quickly does what we eat affect the d13C of our natural compounds? We make new testosterone every day, and metabolize it fairly rapidly (within a couple of days) (in an hour or less [13]). So does the new stuff come from last night's meal? Or longer term bodily carbon stores?

This question has now been answered! Basically, changes in isotope levels from last night's meal, or even this morning's, can immediately be seen in urinary steroids! [11, 12]

This study was a WADA-sponsored study to examine if dietary changes could lead to false positives on the isotope test. The scientists who were awarded the study, and are two of it's authors, are U. Flenker and W. Schanzer. Google these guys, and you'll find they're everywhere in testosterone isotope literature. The point is that this is hardly some fringe study. It was sponsored by WADA and run by highly referenced experts in this very specific field.

Well, let me be honest. They found that diet did not affect isotope ratios to the three per mil standard used in this test (I've now found a pdf file of a presentation of the results - there's a detailed analysis here). But what else did they find? They found that some subjects showed isotope changes in their testosterone from dietary changes essentially immediately (they were having their urine sampled five times per day). They also found that this varied by subject. The study wasn't just of testosterone, but of several urinary steroids, and the study also found that the other steroids were also quickly affected by diet. But they found that the speed of this effect varied depending on the steroid.

This is critically important, because it has been reported that the WADA test compares the testosterone isotope level to that of another steroid. If the other steroid lasts a lot longer in the body, then it's isotope level might take a day or more to catch up with sudden changes in the isotope level of testosterone. This would lead to a failed test.

But still, they found that diet couldn't lead to a false positive, right? Well maybe. Again, I don't have the exact numbers from the tests yet. But their test consisted of six subjects, who were not competing in the Tour de France, not experiencing surges in testosterone production from atletic competition, and not consuming eight to ten thousand calories per day.

One thing that seems very likely is that Floyd experienced a natural testosterone surge, as a result of his intense competitive emotions, and also his victory that day [8, 9]. This could (and perhaps even probably would) have the affect of causing his testosterone d13C value to diverge from any baseline steroid that did not surge in production at the same time, because both steroids would be mixing in the body with steroids that were produced yesterday, but had not yet metabolised.

Cholesterol and Testosterone

Testosterone is synthesized in the body from cholesterol [14]. Cholesterol is in the food we eat, but it is also made by the liver [15], and in fact almost all of the cholesterol we use comes from the liver rather than directly from food [16].

The particularly interesting thing here is that alcohol is broken down by the liver. It seems likely that this means that the carbon in the alcohol we drink will end up very rapidly in the cholesterol we produce, and hence the testosterone. Not only that, but the consumption of alcohol can boost testosterone, which would tend to exagerate any changes on d13C caused by the absorption of alcohol [17]. (Although this is only a short-term effect -- long term alchohol abuse depresses testosterone production.) Some questions that this raises are, what's the half-life of cholesterol under normal circumstances. And, if you bonk, is cholesterol depleted?

So one more little tidbit. Alcohol is absorbed directly into the bloodstream, where it is cleared out by the liver, which converts it into acetic acid. And cholesterol in the liver is synthesized from acetic acid. [reference missing] It looks quite likely that alcohol can have more rapid skewing of carbon-13 values in testosterone than other foods we ingest might have.

So what's the d13C of the beer and the whiskey. Well, the fermentation process does tend to reduce the carbon-13 (makes the d13C more negative), by perhaps a couple per mil. If the organic matter being fermented came from C3 plants, you can end up with alcohol that has a d13C of -29 per mil [18]. Beer is made exclusively from C3 plants, and you could then conclude that drinking beer might have a rapid and significant impact on the carbon-13 ratios which could lead to a false positive.

What about the whiskey? Well, that one cuts the other way. American bourbons and whiskies must have at least 51% corn in their mash. This means that at least half of the mash is going to have a d13C of around -13 per mil. Even if fermentation makes that a bit more negative, then the average of half corn mash and half C4 plants could at best (for Floyd) produce alcohol with a d13C of around -23 per mil, which would pull him away from a false postive.

Side comment on the media

"It's almost like a fingerprint, as the experts have explained it to us - in actually finding a specific carbon atom that is only present in synthetic testosterone, and comparing that to natural testosterone."

I've looked into this alleged foolproof test that can tell the difference between natural and synthetic testosterone, and it seems to have a single source: Jacques De Ceaurriz, head of the Chatenay-Malabry laboratory. He's the same head that assured the world that the tests that showed that Lance doped were completely reliable, and which were later claimed to be completely worthless by UCI's investigator. Here's what he said about the isotope test:

"It's foolproof. This analysis tells the difference between endogenous and exogenous. No error is possible in isotopic readings."

I don't know what his background is, and I don't know what his agenda is, and I don't know what he might be smoking. But let me say again, it isn't possible to tell the difference between endogenous and exogenous testosterone, in any way that could be reasonably be characterized as foolproof or without error. You can only measure variations in carbon-13 ratios, and then say based on the variation that it is more or less likely to include an exogenous source.

The effect of his untrue statement can not be underestimated. This quote was used in numerous articles, and within a day the headlines were that synthetic testosterone had been "found" or "detected" in Landis' urine. The reports would typically go on to say in one way or another that this was nearly ironclad proof of his guilt. And so began the public destruction of Floyd.

There are other experts out there saying other things. But they are generally less specific. They don't use words like "find", "detect", "differentiate", or "foolproof". They only comment that the test is reliable and accurate, which in the absence of the "foolproof" language would leave open questions of how reliable and how accurate. Or they simply refer to other experts, as Sanjay Gupta did.

Open questions: